[09/03 13:23:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 13:23:30] detectron2 INFO: Environment info:
----------------------  ---------------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.0 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     True
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.10.0+cu102 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  ---------------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 13:23:30] detectron2 INFO: Command line arguments: None
[09/03 13:23:30] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 13:23:30] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 13:23:30] d2.utils.env INFO: Using a generated random seed 30833344
[09/03 13:23:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:23:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:23:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:23:36] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 13:24:56] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 13:24:57] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.0 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     True
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.0 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  /data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision/_C.so
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 13:24:57] detectron2 INFO: Command line arguments: None
[09/03 13:24:57] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 13:24:57] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 13:24:57] d2.utils.env INFO: Using a generated random seed 57093257
[09/03 13:24:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:24:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:24:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:25:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 13:39:32] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 13:39:33] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 13:39:33] detectron2 INFO: Command line arguments: None
[09/03 13:39:33] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 13:39:33] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 13:39:33] d2.utils.env INFO: Using a generated random seed 33525286
[09/03 13:39:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:39:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:39:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 13:39:38] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:20:52] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:20:52] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:20:52] detectron2 INFO: Command line arguments: None
[09/03 14:20:52] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:20:52] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:20:52] d2.utils.env INFO: Using a generated random seed 52827461
[09/03 14:20:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:20:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:20:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:20:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:46:54] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:46:55] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:46:55] detectron2 INFO: Command line arguments: None
[09/03 14:46:55] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:46:55] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:46:55] d2.utils.env INFO: Using a generated random seed 55125752
[09/03 14:46:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:46:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:46:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:01] detectron2 INFO: Command line arguments: None
[09/03 14:47:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:01] d2.utils.env INFO: Using a generated random seed 1738717
[09/03 14:47:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:03] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:04] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:04] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:04] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:04] detectron2 INFO: Command line arguments: None
[09/03 14:47:04] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:04] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:04] d2.utils.env INFO: Using a generated random seed 4547919
[09/03 14:47:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:05] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:06] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:06] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:07] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:07] detectron2 INFO: Command line arguments: None
[09/03 14:47:07] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:07] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:07] d2.utils.env INFO: Using a generated random seed 7189850
[09/03 14:47:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:08] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:09] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:09] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:09] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:09] detectron2 INFO: Command line arguments: None
[09/03 14:47:09] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:09] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:09] d2.utils.env INFO: Using a generated random seed 9768793
[09/03 14:47:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:11] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:11] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:11] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:12] detectron2 INFO: Command line arguments: None
[09/03 14:47:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:12] d2.utils.env INFO: Using a generated random seed 12287705
[09/03 14:47:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:13] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:14] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:14] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:14] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:14] detectron2 INFO: Command line arguments: None
[09/03 14:47:14] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:14] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:14] d2.utils.env INFO: Using a generated random seed 14892963
[09/03 14:47:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:17] detectron2 INFO: Command line arguments: None
[09/03 14:47:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:17] d2.utils.env INFO: Using a generated random seed 17559912
[09/03 14:47:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:18] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:19] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:19] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:20] detectron2 INFO: Command line arguments: None
[09/03 14:47:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:20] d2.utils.env INFO: Using a generated random seed 20222714
[09/03 14:47:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:22] detectron2 INFO: Command line arguments: None
[09/03 14:47:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:22] d2.utils.env INFO: Using a generated random seed 22830637
[09/03 14:47:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:24] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:25] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:25] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:25] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:25] detectron2 INFO: Command line arguments: None
[09/03 14:47:25] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:25] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:25] d2.utils.env INFO: Using a generated random seed 25435980
[09/03 14:47:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:27] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:27] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:28] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:28] detectron2 INFO: Command line arguments: None
[09/03 14:47:28] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:28] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:28] d2.utils.env INFO: Using a generated random seed 28148566
[09/03 14:47:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:29] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:30] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:30] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:30] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:30] detectron2 INFO: Command line arguments: None
[09/03 14:47:30] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:30] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:30] d2.utils.env INFO: Using a generated random seed 30802608
[09/03 14:47:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:32] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:33] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:33] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:33] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:33] detectron2 INFO: Command line arguments: None
[09/03 14:47:33] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:33] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:33] d2.utils.env INFO: Using a generated random seed 33420420
[09/03 14:47:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:34] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:35] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:35] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:35] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:35] detectron2 INFO: Command line arguments: None
[09/03 14:47:35] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:35] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:35] d2.utils.env INFO: Using a generated random seed 35986645
[09/03 14:47:35] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:35] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:35] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:37] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:38] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:38] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:38] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:38] detectron2 INFO: Command line arguments: None
[09/03 14:47:38] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:38] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:38] d2.utils.env INFO: Using a generated random seed 38602721
[09/03 14:47:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:39] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:40] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:40] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:41] detectron2 INFO: Command line arguments: None
[09/03 14:47:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:41] d2.utils.env INFO: Using a generated random seed 41144795
[09/03 14:47:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:42] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:43] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:43] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:43] detectron2 INFO: Command line arguments: None
[09/03 14:47:43] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:43] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:43] d2.utils.env INFO: Using a generated random seed 43782597
[09/03 14:47:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:46] detectron2 INFO: Command line arguments: None
[09/03 14:47:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:46] d2.utils.env INFO: Using a generated random seed 46342126
[09/03 14:47:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:47] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:48] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:48] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:48] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:48] detectron2 INFO: Command line arguments: None
[09/03 14:47:48] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:48] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:48] d2.utils.env INFO: Using a generated random seed 48974041
[09/03 14:47:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:51] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:51] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:51] detectron2 INFO: Command line arguments: None
[09/03 14:47:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:51] d2.utils.env INFO: Using a generated random seed 51594846
[09/03 14:47:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:53] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:53] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:54] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:54] detectron2 INFO: Command line arguments: None
[09/03 14:47:54] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:54] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:54] d2.utils.env INFO: Using a generated random seed 54207005
[09/03 14:47:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:56] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:56] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:56] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:56] detectron2 INFO: Command line arguments: None
[09/03 14:47:56] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:56] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:56] d2.utils.env INFO: Using a generated random seed 56888546
[09/03 14:47:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:58] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:47:58] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:47:58] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:47:59] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:47:59] detectron2 INFO: Command line arguments: None
[09/03 14:47:59] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:47:59] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:47:59] d2.utils.env INFO: Using a generated random seed 59276573
[09/03 14:47:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:47:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:01] detectron2 INFO: Command line arguments: None
[09/03 14:48:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:01] d2.utils.env INFO: Using a generated random seed 1876075
[09/03 14:48:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:03] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:04] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:04] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:04] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:04] detectron2 INFO: Command line arguments: None
[09/03 14:48:04] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:04] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:04] d2.utils.env INFO: Using a generated random seed 4569237
[09/03 14:48:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:05] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:06] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:06] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:07] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:07] detectron2 INFO: Command line arguments: None
[09/03 14:48:07] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:07] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:07] d2.utils.env INFO: Using a generated random seed 7142131
[09/03 14:48:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:08] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:09] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:09] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:09] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:09] detectron2 INFO: Command line arguments: None
[09/03 14:48:09] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:09] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:09] d2.utils.env INFO: Using a generated random seed 9822753
[09/03 14:48:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:11] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:12] detectron2 INFO: Command line arguments: None
[09/03 14:48:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:12] d2.utils.env INFO: Using a generated random seed 12486850
[09/03 14:48:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:13] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:14] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:14] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:15] detectron2 INFO: Command line arguments: None
[09/03 14:48:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:15] d2.utils.env INFO: Using a generated random seed 15203276
[09/03 14:48:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:17] detectron2 INFO: Command line arguments: None
[09/03 14:48:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:17] d2.utils.env INFO: Using a generated random seed 17859941
[09/03 14:48:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:20] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:20] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:20] detectron2 INFO: Command line arguments: None
[09/03 14:48:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:20] d2.utils.env INFO: Using a generated random seed 20548706
[09/03 14:48:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:23] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:23] detectron2 INFO: Command line arguments: None
[09/03 14:48:23] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:23] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:23] d2.utils.env INFO: Using a generated random seed 23304932
[09/03 14:48:23] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:23] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:23] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:24] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:25] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:25] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:25] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:25] detectron2 INFO: Command line arguments: None
[09/03 14:48:25] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:25] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:25] d2.utils.env INFO: Using a generated random seed 25830505
[09/03 14:48:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:27] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:28] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:28] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:28] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:28] detectron2 INFO: Command line arguments: None
[09/03 14:48:28] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:28] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:28] d2.utils.env INFO: Using a generated random seed 28458302
[09/03 14:48:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:29] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:30] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:30] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:31] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:31] detectron2 INFO: Command line arguments: None
[09/03 14:48:31] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:31] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:31] d2.utils.env INFO: Using a generated random seed 31127457
[09/03 14:48:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:32] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:33] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:33] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:33] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:33] detectron2 INFO: Command line arguments: None
[09/03 14:48:33] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:33] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:33] d2.utils.env INFO: Using a generated random seed 33812024
[09/03 14:48:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:35] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:36] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:36] detectron2 INFO: Command line arguments: None
[09/03 14:48:36] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:36] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:36] d2.utils.env INFO: Using a generated random seed 36452534
[09/03 14:48:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:37] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:38] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:38] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:39] detectron2 INFO: Command line arguments: None
[09/03 14:48:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:39] d2.utils.env INFO: Using a generated random seed 39161734
[09/03 14:48:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:41] detectron2 INFO: Command line arguments: None
[09/03 14:48:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:41] d2.utils.env INFO: Using a generated random seed 41776937
[09/03 14:48:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:44] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:44] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:44] detectron2 INFO: Command line arguments: None
[09/03 14:48:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:44] d2.utils.env INFO: Using a generated random seed 44418747
[09/03 14:48:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:47] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:47] detectron2 INFO: Command line arguments: None
[09/03 14:48:47] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:47] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:47] d2.utils.env INFO: Using a generated random seed 47069540
[09/03 14:48:47] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:47] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:47] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:48] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:49] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:49] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:49] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:49] detectron2 INFO: Command line arguments: None
[09/03 14:48:49] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:49] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:49] d2.utils.env INFO: Using a generated random seed 49717310
[09/03 14:48:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:51] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:52] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:52] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:52] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:52] detectron2 INFO: Command line arguments: None
[09/03 14:48:52] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:52] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:52] d2.utils.env INFO: Using a generated random seed 52405370
[09/03 14:48:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:53] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:54] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:54] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:54] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:54] detectron2 INFO: Command line arguments: None
[09/03 14:48:54] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:54] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:54] d2.utils.env INFO: Using a generated random seed 54934270
[09/03 14:48:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:56] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:57] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:57] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:48:57] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:48:57] detectron2 INFO: Command line arguments: None
[09/03 14:48:57] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:48:57] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:48:57] d2.utils.env INFO: Using a generated random seed 57649874
[09/03 14:48:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:48:58] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:48:59] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:48:59] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:00] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:00] detectron2 INFO: Command line arguments: None
[09/03 14:49:00] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:00] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:00] d2.utils.env INFO: Using a generated random seed 211225
[09/03 14:49:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:01] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:02] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:02] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:02] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:02] detectron2 INFO: Command line arguments: None
[09/03 14:49:02] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:02] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:02] d2.utils.env INFO: Using a generated random seed 2985949
[09/03 14:49:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:04] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:05] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:05] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:05] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:05] detectron2 INFO: Command line arguments: None
[09/03 14:49:05] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:05] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:05] d2.utils.env INFO: Using a generated random seed 5722372
[09/03 14:49:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:06] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:08] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:08] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:08] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:08] detectron2 INFO: Command line arguments: None
[09/03 14:49:08] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:08] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:08] d2.utils.env INFO: Using a generated random seed 8306794
[09/03 14:49:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:09] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:10] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:10] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:10] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:10] detectron2 INFO: Command line arguments: None
[09/03 14:49:10] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:10] d2.utils.env INFO: Using a generated random seed 11004840
[09/03 14:49:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:12] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:13] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:13] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:13] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:13] detectron2 INFO: Command line arguments: None
[09/03 14:49:13] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:13] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:13] d2.utils.env INFO: Using a generated random seed 13744644
[09/03 14:49:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:15] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:16] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:16] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:16] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:16] detectron2 INFO: Command line arguments: None
[09/03 14:49:16] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:16] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:16] d2.utils.env INFO: Using a generated random seed 16409548
[09/03 14:49:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:17] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:18] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:18] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:19] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:19] detectron2 INFO: Command line arguments: None
[09/03 14:49:19] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:19] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:19] d2.utils.env INFO: Using a generated random seed 19053610
[09/03 14:49:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:20] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:21] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:21] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:21] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:21] detectron2 INFO: Command line arguments: None
[09/03 14:49:21] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:21] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:21] d2.utils.env INFO: Using a generated random seed 21681236
[09/03 14:49:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:22] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:24] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:24] detectron2 INFO: Command line arguments: None
[09/03 14:49:24] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:24] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:24] d2.utils.env INFO: Using a generated random seed 24314434
[09/03 14:49:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:25] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:26] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:27] detectron2 INFO: Command line arguments: None
[09/03 14:49:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:27] d2.utils.env INFO: Using a generated random seed 27041952
[09/03 14:49:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:29] detectron2 INFO: Command line arguments: None
[09/03 14:49:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:29] d2.utils.env INFO: Using a generated random seed 29631286
[09/03 14:49:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:30] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:32] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:32] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:32] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:32] detectron2 INFO: Command line arguments: None
[09/03 14:49:32] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:32] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:32] d2.utils.env INFO: Using a generated random seed 32332739
[09/03 14:49:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:34] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:34] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:34] detectron2 INFO: Command line arguments: None
[09/03 14:49:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:34] d2.utils.env INFO: Using a generated random seed 34979304
[09/03 14:49:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:36] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:37] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:37] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:37] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:37] detectron2 INFO: Command line arguments: None
[09/03 14:49:37] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:37] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:37] d2.utils.env INFO: Using a generated random seed 37536469
[09/03 14:49:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:38] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:39] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:39] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:40] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:40] detectron2 INFO: Command line arguments: None
[09/03 14:49:40] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:40] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:40] d2.utils.env INFO: Using a generated random seed 40126679
[09/03 14:49:40] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:40] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:40] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:41] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:42] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:42] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:42] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:42] detectron2 INFO: Command line arguments: None
[09/03 14:49:42] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:42] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:42] d2.utils.env INFO: Using a generated random seed 42809433
[09/03 14:49:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:44] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:45] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:45] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:45] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:45] detectron2 INFO: Command line arguments: None
[09/03 14:49:45] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:45] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:45] d2.utils.env INFO: Using a generated random seed 45467401
[09/03 14:49:45] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:45] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:45] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:46] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:47] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:47] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:48] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:48] detectron2 INFO: Command line arguments: None
[09/03 14:49:48] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:48] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:48] d2.utils.env INFO: Using a generated random seed 48129706
[09/03 14:49:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:49] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:50] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:50] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:50] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:50] detectron2 INFO: Command line arguments: None
[09/03 14:49:50] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:50] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:50] d2.utils.env INFO: Using a generated random seed 50737990
[09/03 14:49:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:51] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:53] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:53] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:53] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:53] detectron2 INFO: Command line arguments: None
[09/03 14:49:53] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:53] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:53] d2.utils.env INFO: Using a generated random seed 53425336
[09/03 14:49:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:54] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:55] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:55] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:55] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:55] detectron2 INFO: Command line arguments: None
[09/03 14:49:55] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:55] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:55] d2.utils.env INFO: Using a generated random seed 56001670
[09/03 14:49:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:49:58] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:49:58] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:49:58] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:49:58] detectron2 INFO: Command line arguments: None
[09/03 14:49:58] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:49:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:49:58] d2.utils.env INFO: Using a generated random seed 58657946
[09/03 14:49:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:49:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:01] detectron2 INFO: Command line arguments: None
[09/03 14:50:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:01] d2.utils.env INFO: Using a generated random seed 1385208
[09/03 14:50:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:03] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:03] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:03] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:03] detectron2 INFO: Command line arguments: None
[09/03 14:50:03] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:03] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:03] d2.utils.env INFO: Using a generated random seed 4068109
[09/03 14:50:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:05] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:06] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:06] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:06] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:06] detectron2 INFO: Command line arguments: None
[09/03 14:50:06] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:06] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:06] d2.utils.env INFO: Using a generated random seed 6673187
[09/03 14:50:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:07] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:08] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:08] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:09] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:09] detectron2 INFO: Command line arguments: None
[09/03 14:50:09] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:09] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:09] d2.utils.env INFO: Using a generated random seed 9214287
[09/03 14:50:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:10] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:11] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:11] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:11] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:11] detectron2 INFO: Command line arguments: None
[09/03 14:50:11] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:11] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:11] d2.utils.env INFO: Using a generated random seed 11891377
[09/03 14:50:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:13] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:14] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:14] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:14] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:14] detectron2 INFO: Command line arguments: None
[09/03 14:50:14] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:14] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:14] d2.utils.env INFO: Using a generated random seed 14575441
[09/03 14:50:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:15] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:16] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:16] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:17] detectron2 INFO: Command line arguments: None
[09/03 14:50:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:17] d2.utils.env INFO: Using a generated random seed 17206823
[09/03 14:50:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:18] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:19] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:19] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:19] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:19] detectron2 INFO: Command line arguments: None
[09/03 14:50:19] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:19] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:19] d2.utils.env INFO: Using a generated random seed 19840545
[09/03 14:50:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:22] detectron2 INFO: Command line arguments: None
[09/03 14:50:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:22] d2.utils.env INFO: Using a generated random seed 22442997
[09/03 14:50:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:23] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:25] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:25] detectron2 INFO: Command line arguments: None
[09/03 14:50:25] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:25] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:25] d2.utils.env INFO: Using a generated random seed 25107402
[09/03 14:50:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:27] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:27] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:27] detectron2 INFO: Command line arguments: None
[09/03 14:50:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:27] d2.utils.env INFO: Using a generated random seed 27776725
[09/03 14:50:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:29] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:30] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:30] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:30] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:30] detectron2 INFO: Command line arguments: None
[09/03 14:50:30] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:30] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:30] d2.utils.env INFO: Using a generated random seed 30412638
[09/03 14:50:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:31] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:32] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:32] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:33] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:33] detectron2 INFO: Command line arguments: None
[09/03 14:50:33] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:33] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:33] d2.utils.env INFO: Using a generated random seed 33069036
[09/03 14:50:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:34] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:35] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:35] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:35] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:35] detectron2 INFO: Command line arguments: None
[09/03 14:50:35] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:35] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:35] d2.utils.env INFO: Using a generated random seed 35629625
[09/03 14:50:35] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:35] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:35] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:36] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:38] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:38] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:38] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:38] detectron2 INFO: Command line arguments: None
[09/03 14:50:38] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:38] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:38] d2.utils.env INFO: Using a generated random seed 38355663
[09/03 14:50:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:39] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:40] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:40] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:40] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:40] detectron2 INFO: Command line arguments: None
[09/03 14:50:40] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:40] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:40] d2.utils.env INFO: Using a generated random seed 41047347
[09/03 14:50:40] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:42] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:43] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:43] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:43] detectron2 INFO: Command line arguments: None
[09/03 14:50:43] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:43] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:43] d2.utils.env INFO: Using a generated random seed 43656752
[09/03 14:50:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:44] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:45] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:45] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:46] detectron2 INFO: Command line arguments: None
[09/03 14:50:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:46] d2.utils.env INFO: Using a generated random seed 46247615
[09/03 14:50:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:47] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:48] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:48] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:48] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:48] detectron2 INFO: Command line arguments: None
[09/03 14:50:48] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:48] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:48] d2.utils.env INFO: Using a generated random seed 48944079
[09/03 14:50:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:51] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:51] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:51] detectron2 INFO: Command line arguments: None
[09/03 14:50:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:51] d2.utils.env INFO: Using a generated random seed 51600453
[09/03 14:50:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:54] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:54] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:54] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:54] detectron2 INFO: Command line arguments: None
[09/03 14:50:54] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:54] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:54] d2.utils.env INFO: Using a generated random seed 54277258
[09/03 14:50:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:56] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:56] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:56] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:56] detectron2 INFO: Command line arguments: None
[09/03 14:50:56] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:56] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:56] d2.utils.env INFO: Using a generated random seed 56905843
[09/03 14:50:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:58] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:50:59] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:50:59] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:50:59] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:50:59] detectron2 INFO: Command line arguments: None
[09/03 14:50:59] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:50:59] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:50:59] d2.utils.env INFO: Using a generated random seed 59477861
[09/03 14:50:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:50:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:02] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:02] detectron2 INFO: Command line arguments: None
[09/03 14:51:02] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:02] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:02] d2.utils.env INFO: Using a generated random seed 2105731
[09/03 14:51:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:03] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:04] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:04] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:04] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:04] detectron2 INFO: Command line arguments: None
[09/03 14:51:04] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:04] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:04] d2.utils.env INFO: Using a generated random seed 4814378
[09/03 14:51:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:06] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:07] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:07] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:07] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:07] detectron2 INFO: Command line arguments: None
[09/03 14:51:07] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:07] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:07] d2.utils.env INFO: Using a generated random seed 7445837
[09/03 14:51:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:08] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:09] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:09] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:09] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:09] detectron2 INFO: Command line arguments: None
[09/03 14:51:09] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:10] d2.utils.env INFO: Using a generated random seed 10038845
[09/03 14:51:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:11] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:12] detectron2 INFO: Command line arguments: None
[09/03 14:51:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:12] d2.utils.env INFO: Using a generated random seed 12648513
[09/03 14:51:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:13] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:14] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:14] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:15] detectron2 INFO: Command line arguments: None
[09/03 14:51:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:15] d2.utils.env INFO: Using a generated random seed 15254862
[09/03 14:51:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:17] detectron2 INFO: Command line arguments: None
[09/03 14:51:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:17] d2.utils.env INFO: Using a generated random seed 17928798
[09/03 14:51:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/03 14:51:20] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/03 14:51:20] detectron2 INFO: Rank of current process: 0. World size: 1
[09/03 14:51:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.1.0
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/03 14:51:20] detectron2 INFO: Command line arguments: None
[09/03 14:51:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/03 14:51:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/03 14:51:20] d2.utils.env INFO: Using a generated random seed 20484172
[09/03 14:51:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/03 14:51:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:17:58] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:17:59] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:17:59] detectron2 INFO: Command line arguments: None
[09/08 16:17:59] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:17:59] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:17:59] d2.utils.env INFO: Using a generated random seed 59482882
[09/08 16:17:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:17:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:17:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:04] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:05] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:05] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:06] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:06] detectron2 INFO: Command line arguments: None
[09/08 16:18:06] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:06] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:06] d2.utils.env INFO: Using a generated random seed 6126313
[09/08 16:18:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:07] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:08] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:08] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:08] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:08] detectron2 INFO: Command line arguments: None
[09/08 16:18:08] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:08] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:08] d2.utils.env INFO: Using a generated random seed 8540962
[09/08 16:18:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:09] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:10] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:10] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:10] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:10] detectron2 INFO: Command line arguments: None
[09/08 16:18:10] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:10] d2.utils.env INFO: Using a generated random seed 10852680
[09/08 16:18:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:12] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:13] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:13] detectron2 INFO: Command line arguments: None
[09/08 16:18:13] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:13] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:13] d2.utils.env INFO: Using a generated random seed 13182223
[09/08 16:18:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:14] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:15] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:15] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:15] detectron2 INFO: Command line arguments: None
[09/08 16:18:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:15] d2.utils.env INFO: Using a generated random seed 15543562
[09/08 16:18:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:17] detectron2 INFO: Command line arguments: None
[09/08 16:18:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:17] d2.utils.env INFO: Using a generated random seed 17882620
[09/08 16:18:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:19] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:19] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:20] detectron2 INFO: Command line arguments: None
[09/08 16:18:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:20] d2.utils.env INFO: Using a generated random seed 20218291
[09/08 16:18:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:22] detectron2 INFO: Command line arguments: None
[09/08 16:18:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:22] d2.utils.env INFO: Using a generated random seed 22573061
[09/08 16:18:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:23] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:24] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:24] detectron2 INFO: Command line arguments: None
[09/08 16:18:24] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:24] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:24] d2.utils.env INFO: Using a generated random seed 24912167
[09/08 16:18:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:26] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:27] detectron2 INFO: Command line arguments: None
[09/08 16:18:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:27] d2.utils.env INFO: Using a generated random seed 27302366
[09/08 16:18:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:29] detectron2 INFO: Command line arguments: None
[09/08 16:18:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:29] d2.utils.env INFO: Using a generated random seed 29598188
[09/08 16:18:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:30] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:31] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:31] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:31] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:31] detectron2 INFO: Command line arguments: None
[09/08 16:18:31] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:31] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:31] d2.utils.env INFO: Using a generated random seed 31950034
[09/08 16:18:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:34] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:34] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:34] detectron2 INFO: Command line arguments: None
[09/08 16:18:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:34] d2.utils.env INFO: Using a generated random seed 34398855
[09/08 16:18:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:35] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:36] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:36] detectron2 INFO: Command line arguments: None
[09/08 16:18:36] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:36] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:36] d2.utils.env INFO: Using a generated random seed 36968577
[09/08 16:18:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:38] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:39] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:39] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:39] detectron2 INFO: Command line arguments: None
[09/08 16:18:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:39] d2.utils.env INFO: Using a generated random seed 39307591
[09/08 16:18:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:41] detectron2 INFO: Command line arguments: None
[09/08 16:18:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:41] d2.utils.env INFO: Using a generated random seed 41877146
[09/08 16:18:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:43] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:44] detectron2 INFO: Command line arguments: None
[09/08 16:18:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:44] d2.utils.env INFO: Using a generated random seed 44186943
[09/08 16:18:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:46] detectron2 INFO: Command line arguments: None
[09/08 16:18:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:46] d2.utils.env INFO: Using a generated random seed 46621000
[09/08 16:18:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:47] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:49] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:49] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:49] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:49] detectron2 INFO: Command line arguments: None
[09/08 16:18:49] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:49] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:49] d2.utils.env INFO: Using a generated random seed 49335755
[09/08 16:18:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:51] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:51] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:51] detectron2 INFO: Command line arguments: None
[09/08 16:18:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:51] d2.utils.env INFO: Using a generated random seed 51744325
[09/08 16:18:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:53] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:53] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:54] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:54] detectron2 INFO: Command line arguments: None
[09/08 16:18:54] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:54] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:54] d2.utils.env INFO: Using a generated random seed 54056142
[09/08 16:18:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:56] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:56] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:56] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:56] detectron2 INFO: Command line arguments: None
[09/08 16:18:56] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:56] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:56] d2.utils.env INFO: Using a generated random seed 56410840
[09/08 16:18:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:18:58] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:18:58] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:18:58] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:18:58] detectron2 INFO: Command line arguments: None
[09/08 16:18:58] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:18:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:18:58] d2.utils.env INFO: Using a generated random seed 58860805
[09/08 16:18:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:18:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:00] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:00] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:01] detectron2 INFO: Command line arguments: None
[09/08 16:19:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:01] d2.utils.env INFO: Using a generated random seed 1225873
[09/08 16:19:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:03] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:03] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:03] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:03] detectron2 INFO: Command line arguments: None
[09/08 16:19:03] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:03] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:03] d2.utils.env INFO: Using a generated random seed 3506101
[09/08 16:19:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:04] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:05] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:05] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:05] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:05] detectron2 INFO: Command line arguments: None
[09/08 16:19:05] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:05] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:05] d2.utils.env INFO: Using a generated random seed 5852029
[09/08 16:19:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:07] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:07] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:07] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:08] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:08] detectron2 INFO: Command line arguments: None
[09/08 16:19:08] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:08] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:08] d2.utils.env INFO: Using a generated random seed 8174499
[09/08 16:19:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:09] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:10] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:10] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:10] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:10] detectron2 INFO: Command line arguments: None
[09/08 16:19:10] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:10] d2.utils.env INFO: Using a generated random seed 10516341
[09/08 16:19:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:11] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:12] detectron2 INFO: Command line arguments: None
[09/08 16:19:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:12] d2.utils.env INFO: Using a generated random seed 12863666
[09/08 16:19:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:14] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:15] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:15] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:15] detectron2 INFO: Command line arguments: None
[09/08 16:19:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:15] d2.utils.env INFO: Using a generated random seed 15489525
[09/08 16:19:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:17] detectron2 INFO: Command line arguments: None
[09/08 16:19:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:17] d2.utils.env INFO: Using a generated random seed 18016882
[09/08 16:19:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:20] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:20] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:20] detectron2 INFO: Command line arguments: None
[09/08 16:19:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:20] d2.utils.env INFO: Using a generated random seed 20584023
[09/08 16:19:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:22] detectron2 INFO: Command line arguments: None
[09/08 16:19:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:22] d2.utils.env INFO: Using a generated random seed 22979873
[09/08 16:19:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:24] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:25] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:25] detectron2 INFO: Command line arguments: None
[09/08 16:19:25] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:25] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:25] d2.utils.env INFO: Using a generated random seed 25306645
[09/08 16:19:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:27] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:27] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:27] detectron2 INFO: Command line arguments: None
[09/08 16:19:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:27] d2.utils.env INFO: Using a generated random seed 27599964
[09/08 16:19:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:29] detectron2 INFO: Command line arguments: None
[09/08 16:19:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:29] d2.utils.env INFO: Using a generated random seed 29953843
[09/08 16:19:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:31] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:32] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:32] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:32] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:32] detectron2 INFO: Command line arguments: None
[09/08 16:19:32] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:32] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:32] d2.utils.env INFO: Using a generated random seed 32339962
[09/08 16:19:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:34] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:34] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:34] detectron2 INFO: Command line arguments: None
[09/08 16:19:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:34] d2.utils.env INFO: Using a generated random seed 34698908
[09/08 16:19:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:35] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:36] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:36] detectron2 INFO: Command line arguments: None
[09/08 16:19:36] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:36] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:36] d2.utils.env INFO: Using a generated random seed 36999682
[09/08 16:19:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:38] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:39] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:39] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:39] detectron2 INFO: Command line arguments: None
[09/08 16:19:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:39] d2.utils.env INFO: Using a generated random seed 39533539
[09/08 16:19:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:41] detectron2 INFO: Command line arguments: None
[09/08 16:19:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:41] d2.utils.env INFO: Using a generated random seed 41824394
[09/08 16:19:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:43] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:44] detectron2 INFO: Command line arguments: None
[09/08 16:19:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:44] d2.utils.env INFO: Using a generated random seed 44176472
[09/08 16:19:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:46] detectron2 INFO: Command line arguments: None
[09/08 16:19:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:46] d2.utils.env INFO: Using a generated random seed 46527268
[09/08 16:19:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:47] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:48] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:48] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:48] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:48] detectron2 INFO: Command line arguments: None
[09/08 16:19:48] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:48] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:48] d2.utils.env INFO: Using a generated random seed 48946706
[09/08 16:19:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:51] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:51] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:51] detectron2 INFO: Command line arguments: None
[09/08 16:19:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:51] d2.utils.env INFO: Using a generated random seed 51317442
[09/08 16:19:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:53] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:53] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:53] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:53] detectron2 INFO: Command line arguments: None
[09/08 16:19:53] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:53] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:53] d2.utils.env INFO: Using a generated random seed 53844950
[09/08 16:19:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:55] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:55] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:56] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:56] detectron2 INFO: Command line arguments: None
[09/08 16:19:56] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:56] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:56] d2.utils.env INFO: Using a generated random seed 56282710
[09/08 16:19:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:19:58] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:19:58] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:19:58] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:19:58] detectron2 INFO: Command line arguments: None
[09/08 16:19:58] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:19:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:19:58] d2.utils.env INFO: Using a generated random seed 58866820
[09/08 16:19:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:19:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:01] detectron2 INFO: Command line arguments: None
[09/08 16:20:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:01] d2.utils.env INFO: Using a generated random seed 1343303
[09/08 16:20:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:03] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:03] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:03] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:03] detectron2 INFO: Command line arguments: None
[09/08 16:20:03] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:03] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:03] d2.utils.env INFO: Using a generated random seed 3793285
[09/08 16:20:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:05] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:05] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:05] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:05] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:05] detectron2 INFO: Command line arguments: None
[09/08 16:20:05] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:05] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:05] d2.utils.env INFO: Using a generated random seed 6029527
[09/08 16:20:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:07] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:08] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:08] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:08] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:08] detectron2 INFO: Command line arguments: None
[09/08 16:20:08] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:08] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:08] d2.utils.env INFO: Using a generated random seed 8394744
[09/08 16:20:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:09] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:10] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:10] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:10] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:10] detectron2 INFO: Command line arguments: None
[09/08 16:20:10] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:10] d2.utils.env INFO: Using a generated random seed 10759824
[09/08 16:20:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:12] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:12] detectron2 INFO: Command line arguments: None
[09/08 16:20:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:12] d2.utils.env INFO: Using a generated random seed 13057483
[09/08 16:20:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:14] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:15] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:15] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:15] detectron2 INFO: Command line arguments: None
[09/08 16:20:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:15] d2.utils.env INFO: Using a generated random seed 15361993
[09/08 16:20:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:17] detectron2 INFO: Command line arguments: None
[09/08 16:20:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:17] d2.utils.env INFO: Using a generated random seed 17862804
[09/08 16:20:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:20] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:20] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:20] detectron2 INFO: Command line arguments: None
[09/08 16:20:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:20] d2.utils.env INFO: Using a generated random seed 20372617
[09/08 16:20:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:22] detectron2 INFO: Command line arguments: None
[09/08 16:20:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:22] d2.utils.env INFO: Using a generated random seed 22719332
[09/08 16:20:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:23] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:24] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:24] detectron2 INFO: Command line arguments: None
[09/08 16:20:24] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:24] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:24] d2.utils.env INFO: Using a generated random seed 25030526
[09/08 16:20:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:27] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:27] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:27] detectron2 INFO: Command line arguments: None
[09/08 16:20:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:27] d2.utils.env INFO: Using a generated random seed 27376544
[09/08 16:20:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:29] detectron2 INFO: Command line arguments: None
[09/08 16:20:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:29] d2.utils.env INFO: Using a generated random seed 29721654
[09/08 16:20:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:31] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:32] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:32] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:32] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:32] detectron2 INFO: Command line arguments: None
[09/08 16:20:32] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:32] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:32] d2.utils.env INFO: Using a generated random seed 32411072
[09/08 16:20:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:34] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:34] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:34] detectron2 INFO: Command line arguments: None
[09/08 16:20:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:34] d2.utils.env INFO: Using a generated random seed 34825613
[09/08 16:20:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:36] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:37] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:37] detectron2 INFO: Command line arguments: None
[09/08 16:20:37] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:37] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:37] d2.utils.env INFO: Using a generated random seed 37243193
[09/08 16:20:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:38] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:39] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:39] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:39] detectron2 INFO: Command line arguments: None
[09/08 16:20:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:39] d2.utils.env INFO: Using a generated random seed 39588651
[09/08 16:20:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:42] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:42] detectron2 INFO: Command line arguments: None
[09/08 16:20:42] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:42] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:42] d2.utils.env INFO: Using a generated random seed 42105903
[09/08 16:20:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:44] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:44] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:44] detectron2 INFO: Command line arguments: None
[09/08 16:20:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:44] d2.utils.env INFO: Using a generated random seed 44502419
[09/08 16:20:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:46] detectron2 INFO: Command line arguments: None
[09/08 16:20:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:46] d2.utils.env INFO: Using a generated random seed 46939451
[09/08 16:20:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:48] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:49] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:49] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:49] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:49] detectron2 INFO: Command line arguments: None
[09/08 16:20:49] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:49] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:49] d2.utils.env INFO: Using a generated random seed 49302932
[09/08 16:20:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:51] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:51] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:51] detectron2 INFO: Command line arguments: None
[09/08 16:20:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:51] d2.utils.env INFO: Using a generated random seed 51663113
[09/08 16:20:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:54] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:54] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:54] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:54] detectron2 INFO: Command line arguments: None
[09/08 16:20:54] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:54] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:54] d2.utils.env INFO: Using a generated random seed 54335656
[09/08 16:20:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:56] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:56] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:56] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:56] detectron2 INFO: Command line arguments: None
[09/08 16:20:56] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:56] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:56] d2.utils.env INFO: Using a generated random seed 56654700
[09/08 16:20:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:20:59] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:20:59] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:20:59] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:20:59] detectron2 INFO: Command line arguments: None
[09/08 16:20:59] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:20:59] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:20:59] d2.utils.env INFO: Using a generated random seed 59362886
[09/08 16:20:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:20:59] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:01] detectron2 INFO: Command line arguments: None
[09/08 16:21:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:01] d2.utils.env INFO: Using a generated random seed 1957753
[09/08 16:21:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:03] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:04] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:04] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:04] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:04] detectron2 INFO: Command line arguments: None
[09/08 16:21:04] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:04] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:04] d2.utils.env INFO: Using a generated random seed 4560929
[09/08 16:21:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:05] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:06] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:06] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:06] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:06] detectron2 INFO: Command line arguments: None
[09/08 16:21:06] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:06] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:06] d2.utils.env INFO: Using a generated random seed 7002656
[09/08 16:21:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:08] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:09] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:09] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:09] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:09] detectron2 INFO: Command line arguments: None
[09/08 16:21:09] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:09] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:09] d2.utils.env INFO: Using a generated random seed 9280553
[09/08 16:21:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:09] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:10] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:11] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:11] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:11] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:11] detectron2 INFO: Command line arguments: None
[09/08 16:21:11] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:11] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:11] d2.utils.env INFO: Using a generated random seed 11675746
[09/08 16:21:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:12] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:13] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:13] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:13] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:13] detectron2 INFO: Command line arguments: None
[09/08 16:21:13] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:13] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:13] d2.utils.env INFO: Using a generated random seed 14036865
[09/08 16:21:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:15] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:16] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:16] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:16] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:16] detectron2 INFO: Command line arguments: None
[09/08 16:21:16] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:16] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:16] d2.utils.env INFO: Using a generated random seed 16693342
[09/08 16:21:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:17] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:19] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:19] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:19] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:19] detectron2 INFO: Command line arguments: None
[09/08 16:21:19] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:19] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:19] d2.utils.env INFO: Using a generated random seed 19370380
[09/08 16:21:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:20] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:21] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:21] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:21] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:21] detectron2 INFO: Command line arguments: None
[09/08 16:21:21] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:21] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:21] d2.utils.env INFO: Using a generated random seed 22047435
[09/08 16:21:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:23] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/08 16:21:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/08 16:21:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/08 16:21:24] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/08 16:21:24] detectron2 INFO: Command line arguments: None
[09/08 16:21:24] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/08 16:21:24] detectron2 INFO: Full config saved to ./output/config.yaml
[09/08 16:21:24] d2.utils.env INFO: Using a generated random seed 24651220
[09/08 16:21:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/08 16:21:25] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:13] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:20] detectron2 INFO: Command line arguments: None
[09/12 16:14:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:20] d2.utils.env INFO: Using a generated random seed 20540862
[09/12 16:14:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:25] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:26] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:26] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:26] detectron2 INFO: Command line arguments: None
[09/12 16:14:26] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:26] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:26] d2.utils.env INFO: Using a generated random seed 27000631
[09/12 16:14:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:29] detectron2 INFO: Command line arguments: None
[09/12 16:14:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:29] d2.utils.env INFO: Using a generated random seed 29394872
[09/12 16:14:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:30] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:31] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:31] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:31] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:31] detectron2 INFO: Command line arguments: None
[09/12 16:14:31] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:31] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:31] d2.utils.env INFO: Using a generated random seed 31833574
[09/12 16:14:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:33] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:33] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:34] detectron2 INFO: Command line arguments: None
[09/12 16:14:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:34] d2.utils.env INFO: Using a generated random seed 34155468
[09/12 16:14:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:35] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:36] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:36] detectron2 INFO: Command line arguments: None
[09/12 16:14:36] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:36] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:36] d2.utils.env INFO: Using a generated random seed 36533255
[09/12 16:14:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:37] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:38] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:38] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:38] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:38] detectron2 INFO: Command line arguments: None
[09/12 16:14:38] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:38] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:38] d2.utils.env INFO: Using a generated random seed 38849929
[09/12 16:14:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:38] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:40] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:40] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:41] detectron2 INFO: Command line arguments: None
[09/12 16:14:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:41] d2.utils.env INFO: Using a generated random seed 41187110
[09/12 16:14:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:42] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:43] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:43] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:43] detectron2 INFO: Command line arguments: None
[09/12 16:14:43] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:43] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:43] d2.utils.env INFO: Using a generated random seed 43617952
[09/12 16:14:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:44] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:45] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:45] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:45] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:45] detectron2 INFO: Command line arguments: None
[09/12 16:14:45] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:45] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:45] d2.utils.env INFO: Using a generated random seed 45998531
[09/12 16:14:45] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:45] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:45] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:47] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:48] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:48] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:48] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:48] detectron2 INFO: Command line arguments: None
[09/12 16:14:48] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:48] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:48] d2.utils.env INFO: Using a generated random seed 48307474
[09/12 16:14:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:49] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:50] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:50] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:50] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:50] detectron2 INFO: Command line arguments: None
[09/12 16:14:50] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:50] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:50] d2.utils.env INFO: Using a generated random seed 50657327
[09/12 16:14:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:51] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:52] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:52] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:52] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:52] detectron2 INFO: Command line arguments: None
[09/12 16:14:52] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:52] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:52] d2.utils.env INFO: Using a generated random seed 52970988
[09/12 16:14:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:52] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:54] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:55] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:55] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:55] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:55] detectron2 INFO: Command line arguments: None
[09/12 16:14:55] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:55] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:55] d2.utils.env INFO: Using a generated random seed 55407226
[09/12 16:14:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:56] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:14:57] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:14:57] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:14:57] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:14:57] detectron2 INFO: Command line arguments: None
[09/12 16:14:57] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:14:57] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:14:57] d2.utils.env INFO: Using a generated random seed 57987311
[09/12 16:14:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:14:59] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:00] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:00] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:00] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:00] detectron2 INFO: Command line arguments: None
[09/12 16:15:00] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:00] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:00] d2.utils.env INFO: Using a generated random seed 309319
[09/12 16:15:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:01] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:02] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:02] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:02] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:02] detectron2 INFO: Command line arguments: None
[09/12 16:15:02] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:02] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:02] d2.utils.env INFO: Using a generated random seed 2849165
[09/12 16:15:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:02] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:04] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:04] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:04] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:05] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:05] detectron2 INFO: Command line arguments: None
[09/12 16:15:05] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:05] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:05] d2.utils.env INFO: Using a generated random seed 5134967
[09/12 16:15:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:06] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:07] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:07] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:07] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:07] detectron2 INFO: Command line arguments: None
[09/12 16:15:07] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:07] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:07] d2.utils.env INFO: Using a generated random seed 7578475
[09/12 16:15:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:08] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:09] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:09] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:10] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:10] detectron2 INFO: Command line arguments: None
[09/12 16:15:10] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:10] d2.utils.env INFO: Using a generated random seed 10231675
[09/12 16:15:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:11] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:12] detectron2 INFO: Command line arguments: None
[09/12 16:15:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:12] d2.utils.env INFO: Using a generated random seed 12680592
[09/12 16:15:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:13] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:14] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:14] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:15] detectron2 INFO: Command line arguments: None
[09/12 16:15:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:15] d2.utils.env INFO: Using a generated random seed 15095059
[09/12 16:15:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:17] detectron2 INFO: Command line arguments: None
[09/12 16:15:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:17] d2.utils.env INFO: Using a generated random seed 17467929
[09/12 16:15:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:18] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:19] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:19] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:19] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:19] detectron2 INFO: Command line arguments: None
[09/12 16:15:19] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:19] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:19] d2.utils.env INFO: Using a generated random seed 19956009
[09/12 16:15:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:19] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:21] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:21] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:22] detectron2 INFO: Command line arguments: None
[09/12 16:15:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:22] d2.utils.env INFO: Using a generated random seed 22286204
[09/12 16:15:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:23] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:24] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:24] detectron2 INFO: Command line arguments: None
[09/12 16:15:24] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:24] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:24] d2.utils.env INFO: Using a generated random seed 24628533
[09/12 16:15:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:25] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:26] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:26] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:26] detectron2 INFO: Command line arguments: None
[09/12 16:15:26] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:26] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:26] d2.utils.env INFO: Using a generated random seed 26965641
[09/12 16:15:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:28] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:29] detectron2 INFO: Command line arguments: None
[09/12 16:15:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:29] d2.utils.env INFO: Using a generated random seed 29304265
[09/12 16:15:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:30] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:31] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:31] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:31] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:31] detectron2 INFO: Command line arguments: None
[09/12 16:15:31] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:31] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:31] d2.utils.env INFO: Using a generated random seed 31624689
[09/12 16:15:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:32] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:33] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:33] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:33] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:33] detectron2 INFO: Command line arguments: None
[09/12 16:15:33] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:33] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:33] d2.utils.env INFO: Using a generated random seed 34001580
[09/12 16:15:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:35] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:36] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:36] detectron2 INFO: Command line arguments: None
[09/12 16:15:36] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:36] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:36] d2.utils.env INFO: Using a generated random seed 36608972
[09/12 16:15:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:37] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:38] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:38] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:39] detectron2 INFO: Command line arguments: None
[09/12 16:15:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:39] d2.utils.env INFO: Using a generated random seed 39176216
[09/12 16:15:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:41] detectron2 INFO: Command line arguments: None
[09/12 16:15:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:41] d2.utils.env INFO: Using a generated random seed 41711312
[09/12 16:15:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:43] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:44] detectron2 INFO: Command line arguments: None
[09/12 16:15:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:44] d2.utils.env INFO: Using a generated random seed 44137361
[09/12 16:15:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:46] detectron2 INFO: Command line arguments: None
[09/12 16:15:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:46] d2.utils.env INFO: Using a generated random seed 46489263
[09/12 16:15:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:47] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:48] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:48] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:48] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:48] detectron2 INFO: Command line arguments: None
[09/12 16:15:48] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:48] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:48] d2.utils.env INFO: Using a generated random seed 48863955
[09/12 16:15:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:48] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:50] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:50] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:51] detectron2 INFO: Command line arguments: None
[09/12 16:15:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:51] d2.utils.env INFO: Using a generated random seed 51256906
[09/12 16:15:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:53] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:53] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:53] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:53] detectron2 INFO: Command line arguments: None
[09/12 16:15:53] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:53] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:53] d2.utils.env INFO: Using a generated random seed 53574960
[09/12 16:15:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:53] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:54] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:55] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:55] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:55] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:55] detectron2 INFO: Command line arguments: None
[09/12 16:15:55] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:55] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:55] d2.utils.env INFO: Using a generated random seed 55945536
[09/12 16:15:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:55] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:15:57] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:15:57] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:15:58] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:15:58] detectron2 INFO: Command line arguments: None
[09/12 16:15:58] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:15:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:15:58] d2.utils.env INFO: Using a generated random seed 58245915
[09/12 16:15:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:15:59] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:00] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:00] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:00] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:00] detectron2 INFO: Command line arguments: None
[09/12 16:16:00] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:00] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:00] d2.utils.env INFO: Using a generated random seed 811387
[09/12 16:16:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:00] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:02] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:02] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:03] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:03] detectron2 INFO: Command line arguments: None
[09/12 16:16:03] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:03] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:03] d2.utils.env INFO: Using a generated random seed 3172878
[09/12 16:16:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:04] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:05] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:05] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:05] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:05] detectron2 INFO: Command line arguments: None
[09/12 16:16:05] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:05] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:05] d2.utils.env INFO: Using a generated random seed 5594750
[09/12 16:16:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:05] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:06] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:07] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:07] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:07] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:07] detectron2 INFO: Command line arguments: None
[09/12 16:16:07] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:07] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:07] d2.utils.env INFO: Using a generated random seed 8041995
[09/12 16:16:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:07] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:09] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:10] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:10] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:10] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:10] detectron2 INFO: Command line arguments: None
[09/12 16:16:10] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:10] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:10] d2.utils.env INFO: Using a generated random seed 10494230
[09/12 16:16:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:10] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:11] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:12] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:12] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:12] detectron2 INFO: Command line arguments: None
[09/12 16:16:12] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:12] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:12] d2.utils.env INFO: Using a generated random seed 12866002
[09/12 16:16:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:12] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:14] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:15] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:15] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:15] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:15] detectron2 INFO: Command line arguments: None
[09/12 16:16:15] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:15] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:15] d2.utils.env INFO: Using a generated random seed 15369850
[09/12 16:16:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:15] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:16] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:17] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:17] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:17] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:17] detectron2 INFO: Command line arguments: None
[09/12 16:16:17] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:17] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:17] d2.utils.env INFO: Using a generated random seed 17759668
[09/12 16:16:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:17] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:20] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:20] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:20] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:20] detectron2 INFO: Command line arguments: None
[09/12 16:16:20] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:20] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:20] d2.utils.env INFO: Using a generated random seed 20436967
[09/12 16:16:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:20] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:22] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:22] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:22] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:22] detectron2 INFO: Command line arguments: None
[09/12 16:16:22] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:22] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:22] d2.utils.env INFO: Using a generated random seed 22889406
[09/12 16:16:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:22] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:24] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:25] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:25] detectron2 INFO: Command line arguments: None
[09/12 16:16:25] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:25] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:25] d2.utils.env INFO: Using a generated random seed 25280793
[09/12 16:16:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:25] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:27] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:27] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:27] detectron2 INFO: Command line arguments: None
[09/12 16:16:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:27] d2.utils.env INFO: Using a generated random seed 27577802
[09/12 16:16:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:29] detectron2 INFO: Command line arguments: None
[09/12 16:16:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:30] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:30] d2.utils.env INFO: Using a generated random seed 30092183
[09/12 16:16:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:30] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:31] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:32] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:32] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:32] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:32] detectron2 INFO: Command line arguments: None
[09/12 16:16:32] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:32] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:32] d2.utils.env INFO: Using a generated random seed 32464961
[09/12 16:16:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:32] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:34] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:34] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:34] detectron2 INFO: Command line arguments: None
[09/12 16:16:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:34] d2.utils.env INFO: Using a generated random seed 34774928
[09/12 16:16:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:36] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:37] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:37] detectron2 INFO: Command line arguments: None
[09/12 16:16:37] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:37] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:37] d2.utils.env INFO: Using a generated random seed 37140132
[09/12 16:16:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:38] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:39] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:39] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:39] detectron2 INFO: Command line arguments: None
[09/12 16:16:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:39] d2.utils.env INFO: Using a generated random seed 39654042
[09/12 16:16:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:42] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:42] detectron2 INFO: Command line arguments: None
[09/12 16:16:42] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:42] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:42] d2.utils.env INFO: Using a generated random seed 42127393
[09/12 16:16:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:42] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:44] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:44] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:44] detectron2 INFO: Command line arguments: None
[09/12 16:16:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:44] d2.utils.env INFO: Using a generated random seed 44448505
[09/12 16:16:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:46] detectron2 INFO: Command line arguments: None
[09/12 16:16:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:46] d2.utils.env INFO: Using a generated random seed 46800927
[09/12 16:16:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:48] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:48] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:48] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:49] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:49] detectron2 INFO: Command line arguments: None
[09/12 16:16:49] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:49] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:49] d2.utils.env INFO: Using a generated random seed 49155740
[09/12 16:16:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:51] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:51] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:51] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:51] detectron2 INFO: Command line arguments: None
[09/12 16:16:51] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:51] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:51] d2.utils.env INFO: Using a generated random seed 51509970
[09/12 16:16:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:51] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:52] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:53] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:53] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:54] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:54] detectron2 INFO: Command line arguments: None
[09/12 16:16:54] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:54] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:54] d2.utils.env INFO: Using a generated random seed 54164623
[09/12 16:16:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:54] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:56] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:56] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:56] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:56] detectron2 INFO: Command line arguments: None
[09/12 16:16:56] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:56] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:56] d2.utils.env INFO: Using a generated random seed 56597479
[09/12 16:16:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:56] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:57] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:16:58] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:16:58] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:16:58] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:16:58] detectron2 INFO: Command line arguments: None
[09/12 16:16:58] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:16:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:16:58] d2.utils.env INFO: Using a generated random seed 58946281
[09/12 16:16:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:16:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:00] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:01] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:01] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:01] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:01] detectron2 INFO: Command line arguments: None
[09/12 16:17:01] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:01] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:01] d2.utils.env INFO: Using a generated random seed 1351444
[09/12 16:17:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:01] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:03] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:03] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:03] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:03] detectron2 INFO: Command line arguments: None
[09/12 16:17:03] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:03] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:03] d2.utils.env INFO: Using a generated random seed 4059585
[09/12 16:17:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:03] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:04] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:05] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:06] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:06] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:06] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:06] detectron2 INFO: Command line arguments: None
[09/12 16:17:06] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:06] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:06] d2.utils.env INFO: Using a generated random seed 6445150
[09/12 16:17:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:06] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:07] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:08] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:08] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:08] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:08] detectron2 INFO: Command line arguments: None
[09/12 16:17:08] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:08] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:08] d2.utils.env INFO: Using a generated random seed 8974720
[09/12 16:17:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:08] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:10] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:11] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:11] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:11] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:11] detectron2 INFO: Command line arguments: None
[09/12 16:17:11] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:11] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:11] d2.utils.env INFO: Using a generated random seed 11337961
[09/12 16:17:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:11] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:12] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:13] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:13] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:13] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:13] detectron2 INFO: Command line arguments: None
[09/12 16:17:13] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:13] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:13] d2.utils.env INFO: Using a generated random seed 13669011
[09/12 16:17:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:13] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:14] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:16] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:16] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:16] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:16] detectron2 INFO: Command line arguments: None
[09/12 16:17:16] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:16] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:16] d2.utils.env INFO: Using a generated random seed 16372982
[09/12 16:17:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:17] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:18] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:18] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:18] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:18] detectron2 INFO: Command line arguments: None
[09/12 16:17:18] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:18] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:18] d2.utils.env INFO: Using a generated random seed 18694271
[09/12 16:17:18] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:18] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:18] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:21] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:21] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:21] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:21] detectron2 INFO: Command line arguments: None
[09/12 16:17:21] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:21] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:21] d2.utils.env INFO: Using a generated random seed 21404755
[09/12 16:17:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:22] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:24] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:24] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:24] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:24] detectron2 INFO: Command line arguments: None
[09/12 16:17:24] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:24] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:24] d2.utils.env INFO: Using a generated random seed 24383729
[09/12 16:17:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:24] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:25] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:26] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:26] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:26] detectron2 INFO: Command line arguments: None
[09/12 16:17:26] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:26] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:26] d2.utils.env INFO: Using a generated random seed 27011713
[09/12 16:17:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:26] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:28] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:29] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:29] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:29] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:29] detectron2 INFO: Command line arguments: None
[09/12 16:17:29] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:29] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:29] d2.utils.env INFO: Using a generated random seed 29435345
[09/12 16:17:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:30] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:31] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:31] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:31] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:31] detectron2 INFO: Command line arguments: None
[09/12 16:17:31] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:31] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:31] d2.utils.env INFO: Using a generated random seed 31781544
[09/12 16:17:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:31] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:33] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:33] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:34] detectron2 INFO: Command line arguments: None
[09/12 16:17:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:34] d2.utils.env INFO: Using a generated random seed 34101411
[09/12 16:17:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:35] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:36] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:36] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:36] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:36] detectron2 INFO: Command line arguments: None
[09/12 16:17:36] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:36] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:36] d2.utils.env INFO: Using a generated random seed 36498593
[09/12 16:17:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:36] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:37] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:38] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:38] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:39] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:39] detectron2 INFO: Command line arguments: None
[09/12 16:17:39] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:39] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:39] d2.utils.env INFO: Using a generated random seed 39221637
[09/12 16:17:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:39] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:40] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:41] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:41] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:41] detectron2 INFO: Command line arguments: None
[09/12 16:17:41] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:41] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:41] d2.utils.env INFO: Using a generated random seed 41910281
[09/12 16:17:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:41] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:43] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:44] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:44] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:44] detectron2 INFO: Command line arguments: None
[09/12 16:17:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:44] d2.utils.env INFO: Using a generated random seed 44576417
[09/12 16:17:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:45] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:17:46] d2.config.compat WARNING: Config 'configs/bua-caffe/extract-bua-caffe-r101.yaml' has no VERSION. Assuming it to be compatible with latest v2.
[09/12 16:17:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:17:47] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:17:47] detectron2 INFO: Command line arguments: None
[09/12 16:17:47] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:17:47] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:17:47] d2.utils.env INFO: Using a generated random seed 47263625
[09/12 16:17:47] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:47] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:47] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:17:48] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:24:44] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:24:46] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:24:46] detectron2 INFO: Command line arguments: None
[09/12 16:24:46] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:24:46] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:24:46] d2.utils.env INFO: Using a generated random seed 46775874
[09/12 16:24:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:24:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:24:46] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:24:51] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:25:55] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:25:57] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:25:57] detectron2 INFO: Command line arguments: None
[09/12 16:25:57] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:25:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:25:58] d2.utils.env INFO: Using a generated random seed 58096359
[09/12 16:25:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:25:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:25:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:26:03] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:28:35] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:28:37] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:28:37] detectron2 INFO: Command line arguments: None
[09/12 16:28:37] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:28:37] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:28:37] d2.utils.env INFO: Using a generated random seed 37636712
[09/12 16:28:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:28:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:28:37] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:28:42] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 16:29:14] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 16:29:16] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 16:29:16] detectron2 INFO: Command line arguments: None
[09/12 16:29:16] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 16:29:16] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 16:29:16] d2.utils.env INFO: Using a generated random seed 16798131
[09/12 16:29:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:29:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:29:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 16:29:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:29:18] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:29:21] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:29:21] detectron2 INFO: Command line arguments: None
[09/12 21:29:21] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:29:21] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:29:21] d2.utils.env INFO: Using a generated random seed 21161091
[09/12 21:29:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:29:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:29:21] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:29:26] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:29:46] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:29:49] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:29:49] detectron2 INFO: Command line arguments: None
[09/12 21:29:49] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:29:49] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:29:49] d2.utils.env INFO: Using a generated random seed 49489264
[09/12 21:29:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:29:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:29:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:29:54] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:33:12] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:33:14] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:33:14] detectron2 INFO: Command line arguments: None
[09/12 21:33:14] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:33:14] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:33:14] d2.utils.env INFO: Using a generated random seed 14673768
[09/12 21:33:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:33:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:33:14] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:33:19] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:38:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:38:28] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:38:28] detectron2 INFO: Command line arguments: None
[09/12 21:38:28] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:38:28] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:38:28] d2.utils.env INFO: Using a generated random seed 29006570
[09/12 21:38:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:38:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:38:29] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:38:34] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:42:54] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:42:57] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:42:57] detectron2 INFO: Command line arguments: None
[09/12 21:42:57] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:42:57] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:42:57] d2.utils.env INFO: Using a generated random seed 57145417
[09/12 21:42:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:42:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:42:57] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:43:02] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:44:41] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:44:43] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:44:43] detectron2 INFO: Command line arguments: None
[09/12 21:44:43] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:44:43] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:44:43] d2.utils.env INFO: Using a generated random seed 43502337
[09/12 21:44:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:44:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:44:43] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:44:48] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/12 21:53:25] detectron2 INFO: Rank of current process: 0. World size: 1
[09/12 21:53:27] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/12 21:53:27] detectron2 INFO: Command line arguments: None
[09/12 21:53:27] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/12 21:53:27] detectron2 INFO: Full config saved to ./output/config.yaml
[09/12 21:53:27] d2.utils.env INFO: Using a generated random seed 27714527
[09/12 21:53:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:53:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:53:27] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/12 21:53:32] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 10:22:31] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 10:22:33] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 10:22:33] detectron2 INFO: Command line arguments: None
[09/13 10:22:33] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 10:22:33] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 10:22:33] d2.utils.env INFO: Using a generated random seed 33961806
[09/13 10:22:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:22:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:22:33] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:22:39] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 10:36:26] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 10:36:28] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 10:36:28] detectron2 INFO: Command line arguments: None
[09/13 10:36:28] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 10:36:28] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 10:36:28] d2.utils.env INFO: Using a generated random seed 28754642
[09/13 10:36:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:36:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:36:28] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:36:33] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 10:37:43] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 10:37:44] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 10:37:44] detectron2 INFO: Command line arguments: None
[09/13 10:37:44] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 10:37:44] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 10:37:44] d2.utils.env INFO: Using a generated random seed 44975601
[09/13 10:37:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:37:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:37:44] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 10:37:50] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 12:26:47] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 12:26:49] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 12:26:49] detectron2 INFO: Command line arguments: None
[09/13 12:26:49] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 12:26:49] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 12:26:49] d2.utils.env INFO: Using a generated random seed 49709295
[09/13 12:26:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 12:26:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 12:26:49] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 12:26:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 12:27:31] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 12:27:34] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 12:27:34] detectron2 INFO: Command line arguments: None
[09/13 12:27:34] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 12:27:34] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 12:27:34] d2.utils.env INFO: Using a generated random seed 34130381
[09/13 12:27:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 12:27:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 12:27:34] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 12:27:39] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 14:52:49] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 14:52:50] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 14:52:50] detectron2 INFO: Command line arguments: None
[09/13 14:52:50] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 14:52:50] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 14:52:50] d2.utils.env INFO: Using a generated random seed 50133611
[09/13 14:52:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:52:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:52:50] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:52:55] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 14:53:15] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 14:53:16] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 14:53:16] detectron2 INFO: Command line arguments: None
[09/13 14:53:16] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 14:53:16] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 14:53:16] d2.utils.env INFO: Using a generated random seed 16132496
[09/13 14:53:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:53:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:53:16] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:53:21] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
[09/13 14:54:57] detectron2 INFO: Rank of current process: 0. World size: 1
[09/13 14:54:57] detectron2 INFO: Environment info:
----------------------  --------------------------------------------------------------------------
sys.platform            linux
Python                  3.6.5 |Anaconda, Inc.| (default, Apr 29 2018, 16:14:56) [GCC 7.2.0]
numpy                   1.19.5
detectron2              0.3 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/detectron2
Compiler                GCC 7.3
CUDA compiler           CUDA 10.2
detectron2 arch flags   3.7, 5.0, 5.2, 6.0, 6.1, 7.0, 7.5
DETECTRON2_ENV_MODULE   <not set>
PyTorch                 1.7.1 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torch
PyTorch debug build     False
GPU available           True
GPU 0,1,2,3             Tesla V100-SXM2-32GB (arch=7.0)
CUDA_HOME               /usr/local/cuda-10.2
Pillow                  8.3.2
torchvision             0.8.2 @/data/home/sxhong/anaconda3/lib/python3.6/site-packages/torchvision
torchvision arch flags  3.5, 5.0, 6.0, 7.0, 7.5
fvcore                  0.1.5.post20210825
cv2                     3.4.1
----------------------  --------------------------------------------------------------------------
PyTorch built with:
  - GCC 7.3
  - C++ Version: 201402
  - Intel(R) Math Kernel Library Version 2020.0.0 Product Build 20191122 for Intel(R) 64 architecture applications
  - Intel(R) MKL-DNN v1.6.0 (Git Hash 5ef631a030a6f73131c77892041042805a06064f)
  - OpenMP 201511 (a.k.a. OpenMP 4.5)
  - NNPACK is enabled
  - CPU capability usage: AVX2
  - CUDA Runtime 10.2
  - NVCC architecture flags: -gencode;arch=compute_37,code=sm_37;-gencode;arch=compute_50,code=sm_50;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75
  - CuDNN 7.6.5
  - Magma 2.5.2
  - Build settings: BLAS=MKL, BUILD_TYPE=Release, CXX_FLAGS= -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, USE_CUDA=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=OFF, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON, 

[09/13 14:54:57] detectron2 INFO: Command line arguments: None
[09/13 14:54:57] detectron2 INFO: Running with full config:
CUDNN_BENCHMARK: False
DATALOADER:
  ASPECT_RATIO_GROUPING: True
  FILTER_EMPTY_ANNOTATIONS: True
  NUM_WORKERS: 4
  REPEAT_THRESHOLD: 0.0
  SAMPLER_TRAIN: TrainingSampler
DATASETS:
  PRECOMPUTED_PROPOSAL_TOPK_TEST: 1000
  PRECOMPUTED_PROPOSAL_TOPK_TRAIN: 2000
  PROPOSAL_FILES_TEST: ()
  PROPOSAL_FILES_TRAIN: ()
  TEST: ()
  TRAIN: ()
GLOBAL:
  HACK: 1.0
INPUT:
  CROP:
    ENABLED: False
    SIZE: [0.9, 0.9]
    TYPE: relative_range
  FORMAT: BGR
  MASK_FORMAT: polygon
  MAX_SIZE_TEST: 1000
  MAX_SIZE_TRAIN: 1000
  MIN_SIZE_TEST: 600
  MIN_SIZE_TRAIN: (600,)
  MIN_SIZE_TRAIN_SAMPLING: choice
  RANDOM_FLIP: horizontal
MODEL:
  ANCHOR_GENERATOR:
    ANGLES: [[-90, 0, 90]]
    ASPECT_RATIOS: [[0.5, 1.0, 2.0]]
    NAME: DefaultAnchorGenerator
    OFFSET: 0.0
    SIZES: [[4, 8, 16, 32]]
  BACKBONE:
    FREEZE_AT: 3
    NAME: build_bua_resnet_backbone
  BUA:
    ATTRIBUTE:
      NUM_CLASSES: 401
    ATTRIBUTE_ON: True
    CAFFE: True
    EXTRACTOR:
      CONF_THRESH: 0.4
      MAX_BOXES: 45
      MIN_BOXES: 5
      MODE: 2
      OUTPUT_DIR: .output/
    EXTRACT_FEATS: True
    RESNET_VERSION: 1
    RPN:
      CONV_OUT_CHANNELS: 512
  DEVICE: cuda
  FPN:
    FUSE_TYPE: sum
    IN_FEATURES: []
    NORM: 
    OUT_CHANNELS: 256
  KEYPOINT_ON: False
  LOAD_PROPOSALS: False
  MASK_ON: False
  META_ARCHITECTURE: GeneralizedBUARCNN
  PANOPTIC_FPN:
    COMBINE:
      ENABLED: True
      INSTANCES_CONFIDENCE_THRESH: 0.5
      OVERLAP_THRESH: 0.5
      STUFF_AREA_LIMIT: 4096
    INSTANCE_LOSS_WEIGHT: 1.0
  PIXEL_MEAN: [102.9801, 115.9465, 122.7717]
  PIXEL_STD: [1.0, 1.0, 1.0]
  PROPOSAL_GENERATOR:
    MIN_SIZE: 16
    NAME: BUARPN
  RESNETS:
    DEFORM_MODULATED: False
    DEFORM_NUM_GROUPS: 1
    DEFORM_ON_PER_STAGE: [False, False, False, False]
    DEPTH: 101
    NORM: BN
    NUM_GROUPS: 1
    OUT_FEATURES: ['res4']
    RES2_OUT_CHANNELS: 256
    RES5_DILATION: 2
    STEM_OUT_CHANNELS: 64
    STRIDE_IN_1X1: True
    WIDTH_PER_GROUP: 64
  RETINANET:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    FOCAL_LOSS_ALPHA: 0.25
    FOCAL_LOSS_GAMMA: 2.0
    IN_FEATURES: ['p3', 'p4', 'p5', 'p6', 'p7']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.4, 0.5]
    NMS_THRESH_TEST: 0.5
    NORM: 
    NUM_CLASSES: 80
    NUM_CONVS: 4
    PRIOR_PROB: 0.01
    SCORE_THRESH_TEST: 0.05
    SMOOTH_L1_LOSS_BETA: 0.1
    TOPK_CANDIDATES_TEST: 1000
  ROI_BOX_CASCADE_HEAD:
    BBOX_REG_WEIGHTS: ((10.0, 10.0, 5.0, 5.0), (20.0, 20.0, 10.0, 10.0), (30.0, 30.0, 15.0, 15.0))
    IOUS: (0.5, 0.6, 0.7)
  ROI_BOX_HEAD:
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    CLS_AGNOSTIC_BBOX_REG: False
    CONV_DIM: 256
    FC_DIM: 1024
    NAME: 
    NORM: 
    NUM_CONV: 0
    NUM_FC: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIPool
    SMOOTH_L1_BETA: 0.0
    TRAIN_ON_PRED_BOXES: False
  ROI_HEADS:
    BATCH_SIZE_PER_IMAGE: 64
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, 1]
    IOU_THRESHOLDS: [0.5]
    NAME: BUACaffeRes5ROIHeads
    NMS_THRESH_TEST: 0.3
    NUM_CLASSES: 1601
    POSITIVE_FRACTION: 0.5
    PROPOSAL_APPEND_GT: True
    SCORE_THRESH_TEST: -1.0
  ROI_KEYPOINT_HEAD:
    CONV_DIMS: (512, 512, 512, 512, 512, 512, 512, 512)
    LOSS_WEIGHT: 1.0
    MIN_KEYPOINTS_PER_IMAGE: 1
    NAME: KRCNNConvDeconvUpsampleHead
    NORMALIZE_LOSS_BY_VISIBLE_KEYPOINTS: True
    NUM_KEYPOINTS: 17
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  ROI_MASK_HEAD:
    CLS_AGNOSTIC_MASK: False
    CONV_DIM: 256
    NAME: MaskRCNNConvUpsampleHead
    NORM: 
    NUM_CONV: 0
    POOLER_RESOLUTION: 14
    POOLER_SAMPLING_RATIO: 0
    POOLER_TYPE: ROIAlignV2
  RPN:
    BATCH_SIZE_PER_IMAGE: 64
    BBOX_REG_LOSS_TYPE: smooth_l1
    BBOX_REG_LOSS_WEIGHT: 1.0
    BBOX_REG_WEIGHTS: (1.0, 1.0, 1.0, 1.0)
    BOUNDARY_THRESH: -1
    HEAD_NAME: StandardBUARPNHead
    IN_FEATURES: ['res4']
    IOU_LABELS: [0, -1, 1]
    IOU_THRESHOLDS: [0.3, 0.7]
    LOSS_WEIGHT: 1.0
    NMS_THRESH: 0.7
    POSITIVE_FRACTION: 0.5
    POST_NMS_TOPK_TEST: 300
    POST_NMS_TOPK_TRAIN: 2000
    PRE_NMS_TOPK_TEST: 6000
    PRE_NMS_TOPK_TRAIN: 12000
    SMOOTH_L1_BETA: 0.0
  SEM_SEG_HEAD:
    COMMON_STRIDE: 4
    CONVS_DIM: 128
    IGNORE_VALUE: 255
    IN_FEATURES: ['p2', 'p3', 'p4', 'p5']
    LOSS_WEIGHT: 1.0
    NAME: SemSegFPNHead
    NORM: GN
    NUM_CLASSES: 54
  WEIGHTS: weights/bua-caffe-frcn-r101_with_attributes.pth
OUTPUT_DIR: ./output
SEED: -1
SOLVER:
  AMP:
    ENABLED: False
  BASE_LR: 0.001
  BIAS_LR_FACTOR: 1.0
  CHECKPOINT_PERIOD: 5000
  CLIP_GRADIENTS:
    CLIP_TYPE: value
    CLIP_VALUE: 1.0
    ENABLED: False
    NORM_TYPE: 2.0
  GAMMA: 0.1
  IMS_PER_BATCH: 16
  LR_SCHEDULER_NAME: WarmupMultiStepLR
  MAX_ITER: 40000
  MOMENTUM: 0.9
  NESTEROV: False
  REFERENCE_WORLD_SIZE: 0
  STEPS: (30000,)
  WARMUP_FACTOR: 0.001
  WARMUP_ITERS: 1000
  WARMUP_METHOD: linear
  WEIGHT_DECAY: 0.0001
  WEIGHT_DECAY_BIAS: 0.0001
  WEIGHT_DECAY_NORM: 0.0
TEST:
  AUG:
    ENABLED: False
    FLIP: True
    MAX_SIZE: 4000
    MIN_SIZES: (400, 500, 600, 700, 800, 900, 1000, 1100, 1200)
  DETECTIONS_PER_IMAGE: 100
  EVAL_PERIOD: 0
  EXPECTED_RESULTS: []
  KEYPOINT_OKS_SIGMAS: []
  PRECISE_BN:
    ENABLED: False
    NUM_ITER: 200
VERSION: 2
VIS_PERIOD: 0
[09/13 14:54:58] detectron2 INFO: Full config saved to ./output/config.yaml
[09/13 14:54:58] d2.utils.env INFO: Using a generated random seed 58068753
[09/13 14:54:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:54:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:54:58] d2.modeling.backbone.resnet WARNING: ResNet.make_stage(first_stride=) is deprecated!  Use 'stride_per_block' or 'stride' instead.
[09/13 14:55:03] d2.engine.defaults INFO: Model:
GeneralizedBUARCNN(
  (backbone): ResNet(
    (stem): BUABasicStem(
      (conv1): Conv2d(
        3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False
        (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
      )
    )
    (res2): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
        (conv1): Conv2d(
          64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv2): Conv2d(
          64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=64, eps=1e-05)
        )
        (conv3): Conv2d(
          64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=256, eps=1e-05)
        )
      )
    )
    (res3): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
        (conv1): Conv2d(
          256, 128, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv2): Conv2d(
          128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=128, eps=1e-05)
        )
        (conv3): Conv2d(
          128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): FrozenBatchNorm2d(num_features=512, eps=1e-05)
        )
      )
    )
    (res4): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          512, 256, kernel_size=(1, 1), stride=(2, 2), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (3): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (4): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (5): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (6): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (7): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (8): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (9): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (10): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (11): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (12): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (13): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (14): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (15): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (16): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (17): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (18): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (19): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (20): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (21): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (22): BottleneckBlock(
        (conv1): Conv2d(
          1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False
          (norm): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
  )
  (proposal_generator): BUARPN(
    (anchor_generator): DefaultAnchorGenerator(
      (cell_anchors): BufferList()
    )
    (rpn_head): StandardBUARPNHead(
      (conv): Conv2d(1024, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
      (objectness_logits): Conv2d(512, 24, kernel_size=(1, 1), stride=(1, 1))
      (anchor_deltas): Conv2d(512, 48, kernel_size=(1, 1), stride=(1, 1))
    )
  )
  (roi_heads): BUACaffeRes5ROIHeads(
    (pooler): ROIPooler(
      (level_poolers): ModuleList(
        (0): RoIPool(output_size=(14, 14), spatial_scale=0.0625)
      )
    )
    (res5): Sequential(
      (0): BottleneckBlock(
        (shortcut): Conv2d(
          1024, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv1): Conv2d(
          1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (1): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
      (2): BottleneckBlock(
        (conv1): Conv2d(
          2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv2): Conv2d(
          512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(2, 2), dilation=(2, 2), bias=False
          (norm): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
        (conv3): Conv2d(
          512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False
          (norm): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
        )
      )
    )
    (box_predictor): BUACaffeFastRCNNOutputLayers(
      (cls_score): Linear(in_features=2048, out_features=1601, bias=True)
      (bbox_pred): Linear(in_features=2048, out_features=6404, bias=True)
      (cls_embed): Embedding(1601, 256)
      (attr_linear1): Linear(in_features=2304, out_features=512, bias=True)
      (attr_linear2): Linear(in_features=512, out_features=401, bias=True)
    )
  )
)
